CN115618749B - Error compensation method for real-time positioning of large unmanned aerial vehicle - Google Patents
Error compensation method for real-time positioning of large unmanned aerial vehicle Download PDFInfo
- Publication number
- CN115618749B CN115618749B CN202211547468.3A CN202211547468A CN115618749B CN 115618749 B CN115618749 B CN 115618749B CN 202211547468 A CN202211547468 A CN 202211547468A CN 115618749 B CN115618749 B CN 115618749B
- Authority
- CN
- China
- Prior art keywords
- image
- point
- error
- video frame
- aerial vehicle
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000012549 training Methods 0.000 claims abstract description 18
- 230000008569 process Effects 0.000 claims description 11
- 238000012937 correction Methods 0.000 claims description 9
- 230000001419 dependent effect Effects 0.000 claims description 8
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 3
- 206010034719 Personality change Diseases 0.000 claims description 3
- 238000012952 Resampling Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 abstract description 2
- 238000010801 machine learning Methods 0.000 abstract description 2
- 238000013528 artificial neural network Methods 0.000 description 13
- 230000006870 function Effects 0.000 description 5
- 230000002068 genetic effect Effects 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000036544 posture Effects 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Physiology (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Geometry (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Genetics & Genomics (AREA)
- Image Processing (AREA)
Abstract
The invention provides an error compensation method for real-time positioning of a large unmanned aerial vehicle, which comprises the following steps: s1, acquiring an error data set for real-time positioning of an unmanned aerial vehicle; s2, constructing an error model based on the error data set; s3, applying the error model; and S4, updating the error model. The invention fully considers the working mode and working state of each sensor and the problem of data transmission delay, accurately describes the system error of the real-time positioning of the large unmanned aerial vehicle, and avoids the error problem of the matching of the traditional general model and the unmanned aerial vehicle system by combining the sample acquisition and training of the unmanned aerial vehicle in a machine learning mode, thereby remarkably improving the real-time positioning precision of the unmanned aerial vehicle.
Description
Technical Field
The invention relates to the technical field of unmanned aerial vehicle positioning, in particular to an error compensation method for real-time positioning of a large unmanned aerial vehicle.
Background
Real-time positioning of the unmanned aerial vehicle is usually directed by an airborne pos system direct sensor. Correction and correction are needed to be made to the eccentricity component of the on-board GNSS, the eccentricity angle of the inertial navigation and the camera and the azimuth element in the camera in general. Although the above correction can greatly improve the positioning accuracy, the measurement error of inertial navigation, the working state of GNSS \ IMU and the positioning error still restrict the accuracy of direct sensor orientation. These systematic errors are particularly significant when the drone performs positioning tasks for long range, small pitch angles.
Disclosure of Invention
The invention aims to provide an error compensation method for real-time positioning of a large unmanned aerial vehicle, and aims to solve the problem that system errors have great influence when the unmanned aerial vehicle executes positioning tasks of long sight distance and small pitch angle.
The invention provides an error compensation method for real-time positioning of a large unmanned aerial vehicle, which comprises the following steps:
s1, acquiring an error data set for real-time positioning of an unmanned aerial vehicle;
s2, constructing an error model based on the error data set;
s3, applying the error model;
and S4, updating the error model.
Further, step S1 includes the following substeps:
s11, setting a calibration field range, and acquiring a high-resolution image based on a satellite or unmanned aerial vehicle platform in the calibration field range;
s12, planning a flight task of the unmanned aerial vehicle;
s13, enabling the attitude of the photoelectric pod to change periodically in the flight process of the unmanned aerial vehicle;
s14, taking video frames from a video shot by the photoelectric pod camera at fixed intervals, telemetering information of the video frame time and telemetering information 1S before the video frame time, and taking a video frame set I;
s15, for each frame of image in the video frame set I, calculating the geographical position of a pixel in the image according to POS data and executing geographical correction to obtain a new image set J;
s16, respectively carrying out feature matching on the high-resolution images and the images in the new image set J, and endowing the pixels in the video frame set I with accurate geographic coordinates according to the matching result to obtain a point setSum point set;
S17, regarding the ith image in the video frame set I, based on the point setSum point setThe provided pixel coordinates and three-dimensional space coordinates adopt a single-image space back intersection method to calculate exterior orientation elements and line elements thereofIs marked asThe angle element is converted into quaternion and then is recorded as(ii) a Repeating the process, and calculating the exterior orientation element of each image in the video frame set I;
s18, establishing a position error data set and an attitude error data set: for the ith image in the video frame set I, calculating a projection center and a quaternion corrected by an eccentric component and an eccentric angle according to the shooting time and the telemetering information of the previous 1 s:、、and;
will be provided withYaw angle, GNSS operating state and telemetry information solution with respect to the first 1s of a video frameThe yaw angle and the GNSS working state are used as independent variables,establishing a position error data set as a dependent variable; in the same way, willAnd resolving telemetering information of each attitude angle, inertial navigation working state and first 1s of video frameThe attitude angles and the inertial navigation working state are taken as independent variables,establishing an attitude error data set as a dependent variable;
and repeating the process until all the images in the video frame set I are calculated to obtain an error data set.
Further, in step S12, the unmanned aerial vehicle flies around the center O of the calibration yard in the calibration yard by using a petal-shaped flight path.
Further, in step S13, the periodic variation of the attitude of the photovoltaic pod during the flight of the unmanned aerial vehicle means that the azimuth angle of the photovoltaic pod is changedAnd a pitch angleAs a function of time t, expressed as:
in the formula,Tthe period of the attitude change of the photoelectric pod;Tof values not greater than the estimated time of flight of the planned route;To representtIs divided byTThe remainder of (1);andrespectively representing the pitch angle of the photoelectric gondolaMinimum and maximum values of (d);sis a positive integer represented inTIn the period, the pitch angleThe period of time of (a) varies.
Further, step S15 includes:
for the ith image in the video frame set I, directionally calculating an image point set by adopting a direct sensor according to POS dataTo obtain a point set(ii) a Wherein, the image point setW is the width of the image, h is the height of the image;
wherein x and y are point setsThe coordinates of (a); x and Y areThe coordinates of (a); a is 0 、a 1 、a 2 、a 3 、a 4 、a 5 、b 0 、b 1 、b 2 、b 3 、b 4 、b 5 To be a coefficient of undeterminationThe method is obtained by a least square method; calculating the geographic coordinates of each pixel of the ith image in the video frame set I by using a quadratic polynomial model, and then performing gray level resampling to obtain a new image;
and repeating the steps, and processing all the images in the video frame set I to obtain a new image set J.
Further, step S16 includes:
for the ith image in the new image set J, matching a local area of the high-resolution image according to the geographic position of the ith image; performing feature matching by taking the local high-resolution image and the ith image in the new image set J as data sources, screening out a correct matching relation by using a RANSAC algorithm, and respectively marking feature point sets of the image in the high-resolution image and the new image set J as feature point sets、;
Collecting points by formula (3)、Converting the pixel coordinates of the midpoint into geographic coordinates, and recording the geographic coordinates as a point setSum point set:
Wherein A is a homogeneous form of affine transformation matrix of the ith image in the high-resolution image or new image set J, and x and y representOrX and Y represent geographical coordinates;
set pointsSubstituting each point into a quadratic polynomial model to calculate the pixel coordinate of each point in the ith image in the video frame set I, and recording the result as a point set;
For point setsSubstituting the geographical coordinates of each point into a high-precision DEM to calculate the elevation; will point setThe geographical coordinates and the obtained elevation of each point are combined to form three-dimensional space coordinates, and the result is recorded as a point set;
And repeating the steps until all the images in the new image set J are processed.
Further, step S2 includes the following sub-steps:
s21, optimizing a BP neural network with initial weight and threshold value based on a genetic algorithm to serve as an error model;
s22, standardizing the error data set, disorganizing the sequence, and dividing the error data set into a training set and a test set according to a certain proportion;
s23, searching the initial weight and the threshold value of the BP neural network by using a genetic algorithm;
s24, initializing the BP neural network according to the initial weight and the threshold value in the step S23, taking the square sum of the output of the BP neural network and the dependent variable as a loss function, and continuously updating the weight value and the threshold value of the BP neural network by inputting a training set to obtain an error model;
s25, determining the prediction precision, robustness and convergence of the error model according to the performance of the error model on the training set and the test set; if the error model is not good enough, go back to step S21 to adjust the hyperparameter of BP neural network.
Further, step S3 includes the following sub-steps:
s31, calculating the projection center and quaternion of the telemetering information at the current time and the previous 1S time after correction of the eccentric component and the eccentric angle:、、and;
s32, mixingYaw angle, GNSS operating state and telemetry resolution with respect to the first 1s of the video frameInputting the yaw angle and the GNSS working state into a position error model to obtain a projection center compensation value(ii) a In the same way, willCalculation of telemetering information of each attitude angle, inertial navigation working state and first 1s of video frameInputting each attitude angle and inertial navigation working state into an attitude error model to obtain a quaternion compensation value;
S33, calculating the compensated projection center S and quaternion q:
in the formula,in order to compensate for the center of the projection before compensation,in order to compensate for the pre-quaternion,is a quaternion addition method;
and S34, determining a photographic ray equation of any image point according to the projection center S and the quaternion q, wherein the intersection point of the photographic ray equation and the earth surface is the geographic coordinate of the image point.
Further, step S4 includes the following sub-steps:
s41, the calibration task obtains an error data set according to the step S1; judging whether the image principal point coordinates fall in the coverage area of the high-resolution image according to a certain frequency in daily flight, if so, executing steps S14-S18 to obtain an error data set;
and S42, after the number of the new error data sets is accumulated to reach a preset data volume, randomly extracting data with the same data volume as the new error data sets from the original error data sets, combining the extracted data and the new error data sets into a new training set, and updating the original error model in a mode of inputting the new training set into the original error model for training.
In summary, due to the adoption of the technical scheme, the invention has the beneficial effects that:
1. the invention solves the problem of control points by utilizing high-resolution images, only needs to arrange a small number of control points in the early stage, and saves the workload. The position of the image point in the rear intersection is obtained by feature matching, so that the whole process is automated.
2. The error model fully considers the working mode and working state of each sensor and the problem of data transmission delay, accurately describes the system error of the real-time positioning of the large unmanned aerial vehicle, and avoids the error problem of the matching of the traditional general model and the unmanned aerial vehicle system by combining the sample acquisition and training of the specific unmanned aerial vehicle through a machine learning mode, thereby remarkably improving the real-time positioning precision of the unmanned aerial vehicle.
3. The positioning technology has great advantages in emergency search and rescue and battlefield reconnaissance, is flexible and changeable to operate, does not limit the postures of the airplane and the load during ground observation, and can realize real-time high-precision positioning.
4. The invention has simple upgrading and maintenance. Besides the calibration task, the aircraft can accumulate data when performing a daily flight task, and autonomously learn in real time to improve the accuracy of the model.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention, and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
Fig. 1 is a flowchart of an error compensation method for real-time positioning of a large unmanned aerial vehicle in the embodiment of the present invention.
Fig. 2 is a schematic route view of a petal-shaped flight route for flying when planning a flight mission of an unmanned aerial vehicle according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Examples
As shown in fig. 1, the present embodiment provides an error compensation method for real-time positioning of a large unmanned aerial vehicle, including the following steps:
s1, acquiring an error data set of real-time positioning of the unmanned aerial vehicle:
s11, setting a calibration field range, and acquiring a high-resolution image based on a satellite or unmanned aerial vehicle platform in the calibration field range;
s12, planning the flight mission of the unmanned aerial vehicle: respectively get the setsAnd the middle element is used as the flight height H, so that the unmanned aerial vehicle flies around the center O of the calibration field in the calibration field by adopting a petal-shaped flight route.
As shown in fig. 2, the unmanned aerial vehicle starts from the calibration center O, flies to a point a, then turns according to a set turning radius, flies to a point B, and returns to the point O to complete the half 8' flying of the OABO. Then flight in BO direction completes half '8' of OFGO flight. And so on until toThe direction flies back to point O. The complete flight path is OABOFGOCDOOEFOBCOGHODEO.
And S13, enabling the attitude of the photoelectric pod to change periodically in the flight process of the unmanned aerial vehicle.
Specifically, the fact that the attitude of the photoelectric pod is periodically changed in the flight process of the unmanned aerial vehicle means that the azimuth angle of the photoelectric pod is changedAnd a pitch angleAs a function of time t, expressed as:
in the formula,Tthe period of the attitude change of the photoelectric pod;Tof values not greater than the estimated time of flight of the planned route;To representtIs divided byTThe remainder of (1);andrespectively representing the minimum value and the maximum value of the pitch angle of the photoelectric pod;sis a positive integer expressed inTIn the period, the pitch angleThe period of time of (a) varies.
And S14, extracting video frames, telemetering information of the video frame time and telemetering information 1S before the video frame time from the video shot by the photoelectric pod camera at fixed intervals, and extracting a video frame set I.
Specifically, the telemetry information should include the position, attitude angle, electro-optic pod attitude angle, focus information, etc. of the aircraft.
And S15, for each frame of image in the video frame set I, calculating the geographical position of a pixel in the image according to POS data and executing geographical correction to obtain a new image set J.
Specifically, for the ith image in the video frame set I, a direct sensor is adopted to directionally calculate an image point set according to POS dataTo obtain a point set. Wherein,w is the width of the image and h is the height of the image.
wherein x and y are point setsWith X and Y beingA coordinate of (a) 0 、a 1 、a 2 、a 3 、a 4 、a 5 、b 0 、b 1 、b 2 、b 3 、b 4 、b 5 The undetermined coefficient is obtained by a least square method.
And (3) calculating the geographic coordinate of each pixel of the ith image in the video frame set I according to the formula (2), and performing gray level resampling to obtain a new image.
And repeating the steps, and processing all the images in the video frame set I to obtain a new image set J.
S16, respectively carrying out feature matching on the high-resolution images and the images in the new image set J, and endowing the pixels in the video frame set I with accurate geographic coordinates according to the matching result to obtain a point setSum point set。
Specifically, for the ith image in the new image set J, the local area of the high-resolution image is matched according to the geographic position of the ith image. Performing feature matching by taking the local high-resolution image and the ith image in the new image set J as data sources, screening out a correct matching relation by using a RANSAC (Random Sample Consensus) algorithm, and respectively recording the image feature point sets in the high-resolution image and the new image set J as feature point sets、。
The points are collected by the following formula、Converting the pixel coordinate of the midpoint into a geographic coordinate, and recording the geographic coordinate as a point setSum point set:
Wherein A is a homogeneous form of affine transformation matrix of the ith image in the high-resolution image or new image set J, and x and y representOrX, Y represent geographical coordinates.
Will point setSubstituting each point into the formula (2) to calculate the pixel coordinate of the ith image of each point in the video frame set I, and recording the result as a point set。
For point setsAnd (4) substituting the geographical coordinates of each point into a high-precision DEM (Digital Elevation Model) to obtain the Elevation. Will point setThe geographical coordinates and the obtained elevation of each point are combined to form three-dimensional space coordinates, and the result is recorded as a point set。
And repeating the steps until all the images in the new image set J are processed.
S17, for the ith image in the video frame set I, based on the point setSum point setThe provided pixel coordinates and three-dimensional space coordinates are calculated by adopting a single-image space back intersection method, and line elements of the exterior orientation coordinates are recorded asThe angle element is converted into quaternion and then is recorded as. And repeating the process, and calculating the external orientation element of each image in the video frame set I.
And S18, establishing a position error data set and an attitude error data set.
Specifically, for the ith image in the video frame set I, calculating the projection center and quaternion corrected by the eccentric component and the eccentric angle according to the shooting time and the telemetering information of the previous 1 s:、、and。
will be provided withYaw angle, GNSS operating state and telemetry resolution with respect to the first 1s of the video frameThe yaw angle and the GNSS working state are used as independent variables,establishing a position error data set as a dependent variable;in the same way, willCalculation of telemetering information of each attitude angle, inertial navigation working state and first 1s of video frameThe attitude angles and the inertial navigation working state are taken as independent variables,an attitude error data set is established as a dependent variable.
And repeating the process until all the images in the video frame set I are calculated to obtain an error data set.
S2, constructing an error model based on the error data set:
respectively constructing a position error model and an attitude error model by adopting the steps S21-S25 for the position error data set and the attitude error data set:
and S21, using a BP neural network for optimizing initial weight and threshold value based on a genetic algorithm as an error model.
The BP neural network hyper-parameter is set as follows: the number of the hidden layers is one, the initial number of the hidden layers is determined through an empirical formula, and then iterative adjustment is carried out according to results. The activation function is sigmoid, the optimization method is Adam, and the learning rate is 0.01. The number of input layer nodes is determined from the error data set, and the output layer nodes are 3 (position error model) and 4 (attitude error model).
S22, the error data set is standardized and broken, and the error data set is divided into a training set and a testing set according to a certain proportion (such as 7.
And S23, searching the initial weight and the threshold value of the BP neural network by using a genetic algorithm.
And S24, initializing the BP neural network according to the initial weight and the threshold value in the step S23, taking the square sum of the output of the BP neural network and the dependent variable as a loss function, and continuously updating the weight value and the threshold value of the BP neural network by inputting a training set to obtain an error model.
And S25, determining the prediction precision, robustness and convergence of the error model according to the performance of the error model on the training set and the test set. If the error model is not good enough, go back to step S21 to adjust the hyperparameter of BP neural network.
S3, applying the error model:
s31, calculating the projection center and quaternion of the telemetering information at the current time and the previous 1S time after correction of the eccentric component and the eccentric angle:、、and。
s32, mixingYaw angle, GNSS operating state and telemetry information solution with respect to the first 1s of a video frameInputting the yaw angle and the GNSS working state into a position error model to obtain a projection center compensation value(ii) a In the same way, willCalculation of telemetering information of each attitude angle, inertial navigation working state and first 1s of video frameInputting each attitude angle and inertial navigation working state into an attitude error model to obtain a quaternion compensation value。
S33, calculating the compensated projection center S and quaternion q:
in the formula,in order to compensate for the center of the projection before compensation,in order to compensate for the pre-quaternion,is quaternion addition.
And S34, determining a photographing ray equation of any image point according to the projection center S and the quaternion q, wherein the intersection point of the photographing ray equation and the earth surface is the geographic coordinate of the image point.
S4, updating the error model:
s41, the calibration task obtains an error data set according to the step S1; judging whether the image principal point coordinates fall in the coverage area of the high-resolution image according to a certain frequency in daily flight, if so, executing steps S14-S18 to obtain an error data set;
and S42, after the number of the new error data sets is accumulated to reach a preset data volume (such as 1000 error data sets), randomly extracting data with the same data volume as the new error data sets from the original error data sets, combining the extracted data and the new error data sets into a new training set, and updating the original error model in a mode of inputting the new training set into the original error model for training. The model training method is the same as step S21, and the learning rate is adjusted to 0.001.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (5)
1. The error compensation method for real-time positioning of the large unmanned aerial vehicle is characterized by comprising the following steps of:
s1, acquiring an error data set for real-time positioning of an unmanned aerial vehicle;
s2, constructing an error model based on the error data set;
s3, applying the error model;
s4, updating the error model;
step S1 includes the following substeps:
s11, setting a calibration field range, and acquiring a high-resolution image based on a satellite or unmanned aerial vehicle platform in the calibration field range;
s12, planning a flight task of the unmanned aerial vehicle;
s13, enabling the attitude of the photoelectric pod to change periodically in the flight process of the unmanned aerial vehicle;
s14, taking video frames from a video shot by the photoelectric pod camera at fixed intervals, telemetering information of the video frame time and telemetering information 1S before the video frame time, and taking a video frame set I;
s15, for each frame of image in the video frame set I, calculating the geographical position of a pixel in the image according to POS data and executing geographical correction to obtain a new image set J;
s16, respectively carrying out feature matching on the high-resolution images and the images in the new image set J, and endowing the pixels in the video frame set I with accurate geographic coordinates according to the matching result to obtain a point setAnd point set>;
S17, regarding the ith image in the video frame set I, based on the point setAnd point set>The provided pixel coordinate and three-dimensional space coordinate are calculated by adopting a single-image space back intersection method, and the line element is recorded as ^ er>The angle element is converted into quaternion and then recorded as->(ii) a Repeating the process, and calculating the exterior orientation element of each image in the video frame set I;
s18, establishing a position error data set and an attitude error data set: for the ith image in the video frame set I, calculating a projection center and a quaternion corrected by an eccentric component and an eccentric angle according to the shooting time and the telemetering information of the previous 1 s:、、/>and &>;
Will be provided withBased on the measured values of the global navigation system, yaw angle, GNSS operating state and/or telemetry information resolved with respect to the first 1s of the video frame>Based on the deviation angle and the GNSS operating state as arguments>Establishing a position error data set as a dependent variable; in the same way, will->And each attitude angle, inertial navigation working state and telemetering information resolving of 1s before video frame>Each attitude angle and inertial navigation working state are used as independent variables,establishing an attitude error data set as a dependent variable;
repeating the process until all the images in the video frame set I are calculated to obtain an error data set;
step S15 includes:
for the ith image in the video frame set I, directionally calculating an image point set by adopting a direct sensor according to POS dataGet the point set->(ii) a Wherein, the image point set>W is the width of the image, h is the height of the image;
wherein x and y are point setsThe coordinates of (a); x and Y are->The coordinates of (a); a is 0 、a 1 、a 2 、a 3 、a 4 、a 5 、b 0 、b 1 、b 2 、b 3 、b 4 、b 5 The undetermined coefficient is obtained by a least square method; calculating the geographic coordinates of each pixel of the ith image in the video frame set I by using a quadratic polynomial model, and then performing gray level resampling to obtain a new image;
repeating the steps, and processing all the images in the video frame set I to obtain a new image set J;
step S16 includes:
for the ith image in the new image set J, matching a local area of the high-resolution image according to the geographic position of the ith image; performing feature matching by taking the local high-resolution image and the ith image in the new image set J as data sources, screening out a correct matching relation by using a RANSAC algorithm, and respectively marking feature point sets of the image in the high-resolution image and the new image set J as feature point sets、;
By the formula (3) to collect points、/>The pixel coordinate of the midpoint is converted into a geographic coordinate, and the geographic coordinate is recorded as a point set>Sum point set>:
Wherein A is a homogeneous form of affine transformation matrix of the ith image in the high-resolution image or new image set J, and x and y representOr>X and Y represent geographical coordinates;
set pointsSubstituting each point into a quadratic polynomial model to calculate the pixel coordinate of each point in the ith image in the video frame set I, and recording the result as a point set ^ and ^>;
For point setsSubstituting the geographical coordinates of each point into a high-precision DEM to obtain the elevation; collect the point>The geographic coordinate of each point and the obtained elevation are combined to form a three-dimensional space coordinate, and the result is recorded as a point set->;
And repeating the steps until all the images in the new image set J are processed.
2. The method of claim 1, wherein in step S12, the UAV flies around the center O of the calibration center in the calibration field by a petal-shaped flight path.
3. The error compensation method for the real-time positioning of the large unmanned aerial vehicle as claimed in claim 2, wherein in step S13, the step of periodically changing the attitude of the photovoltaic pod during the flight of the unmanned aerial vehicle means that the azimuth angle of the photovoltaic pod is changedAnd pitch angle->As a function of time t, expressed as:
in the formula,Tthe period of the attitude change of the photoelectric pod;Tof values not greater than the estimated time of flight of the planned route;To representtIs divided byTThe remainder of (1); />And &>Respectively representing the minimum value and the maximum value of the pitch angle of the photoelectric pod;sis a positive integer expressed inTIn a cycle, the pitch angle is &>The period of time of (a) varies. />
4. The error compensation method for the real-time positioning of the large unmanned aerial vehicle according to claim 3, wherein the step S3 comprises the following substeps:
s31, calculating the projection center and quaternion of the telemetering information at the current time and the previous 1S time after correction of the eccentric component and the eccentric angle:、/>、/>and &>;
S32, mixingBased on the measured values of the global navigation system, yaw angle, GNSS operating state and/or telemetry information resolved with respect to the first 1s of the video frame>The yaw angle and the GNSS working state are input into the position error model to obtain a projection center compensation value->(ii) a In the same way, will->And each attitude angle, inertial navigation working state and telemetering information resolving of 1s before video frame>Inputting each attitude angle and inertial navigation working state into an attitude error model to obtain a quaternion compensation value->;
S33, calculating the compensated projection center S and quaternion q:
in the formula,in order to centre of projection before compensation>Is a quaternion before compensation>Adding quaternion;
and S34, determining a photographing ray equation of any image point according to the projection center S and the quaternion q, wherein the intersection point of the photographing ray equation and the earth surface is the geographic coordinate of the image point.
5. The error compensation method for the real-time positioning of the large unmanned aerial vehicle according to claim 4, wherein the step S4 comprises the following substeps:
s41, the calibration task obtains an error data set according to the step S1; judging whether the image principal point coordinates fall in the coverage area of the high-resolution image according to a certain frequency in daily flight, if so, executing steps S14-S18 to obtain an error data set;
and S42, after the number of the new error data sets is accumulated to reach a preset data volume, randomly extracting data with the same data volume as the new error data sets from the original error data sets, combining the extracted data and the new error data sets into a new training set, and updating the original error model in a mode of inputting the new training set into the original error model for training.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211547468.3A CN115618749B (en) | 2022-12-05 | 2022-12-05 | Error compensation method for real-time positioning of large unmanned aerial vehicle |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211547468.3A CN115618749B (en) | 2022-12-05 | 2022-12-05 | Error compensation method for real-time positioning of large unmanned aerial vehicle |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115618749A CN115618749A (en) | 2023-01-17 |
CN115618749B true CN115618749B (en) | 2023-04-07 |
Family
ID=84879701
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211547468.3A Active CN115618749B (en) | 2022-12-05 | 2022-12-05 | Error compensation method for real-time positioning of large unmanned aerial vehicle |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115618749B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117169980B (en) * | 2023-11-01 | 2024-01-16 | 中国船舶集团有限公司第七〇七研究所 | Accurate compensation method for gravity measurement acceleration eccentric effect error |
Family Cites Families (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102011077074B4 (en) * | 2011-06-07 | 2022-10-27 | Esg Elektroniksystem- Und Logistik Gmbh | Method and device for determining an error when determining geo-coordinates |
CN103345737B (en) * | 2013-06-04 | 2016-08-10 | 北京航空航天大学 | A kind of UAV high resolution image geometric correction method based on error compensation |
CN105180963B (en) * | 2015-07-22 | 2018-02-16 | 北京航空航天大学 | Unmanned plane telemetry parameter modification method based on online calibration |
CN105222788B (en) * | 2015-09-30 | 2018-07-06 | 清华大学 | The automatic correcting method of the matched aircraft Route Offset error of feature based |
CN105424010A (en) * | 2015-11-17 | 2016-03-23 | 中国人民解放军信息工程大学 | Unmanned aerial vehicle video geographic space information registering method |
CN106127697B (en) * | 2016-06-07 | 2018-12-11 | 哈尔滨工业大学 | EO-1 hyperion geometric correction method is imaged in unmanned aerial vehicle onboard |
CN106352897B (en) * | 2016-08-26 | 2018-06-15 | 杨百川 | It is a kind of based on the silicon MEMS gyro estimation error of monocular vision sensor and bearing calibration |
CN106871927B (en) * | 2017-01-05 | 2020-10-20 | 南京航空航天大学 | Unmanned aerial vehicle photoelectric pod installation error calibration method |
CN110385720B (en) * | 2019-07-26 | 2020-08-04 | 南京航空航天大学 | Robot positioning error compensation method based on deep neural network |
CN110470304B (en) * | 2019-08-19 | 2021-04-20 | 西安因诺航空科技有限公司 | High-precision target positioning and speed measuring method based on unmanned aerial vehicle photoelectric platform |
CN111272196A (en) * | 2020-02-29 | 2020-06-12 | 武汉大学 | In-orbit outside orientation element self-checking and correcting method and system under specific shooting condition |
CN113793270A (en) * | 2021-08-05 | 2021-12-14 | 杭州电子科技大学 | Aerial image geometric correction method based on unmanned aerial vehicle attitude information |
CN114693754B (en) * | 2022-05-30 | 2022-08-19 | 湖南大学 | Unmanned aerial vehicle autonomous positioning method and system based on monocular vision inertial navigation fusion |
CN115187798A (en) * | 2022-06-15 | 2022-10-14 | 中国人民解放军32146部队 | Multi-unmanned aerial vehicle high-precision matching positioning method |
-
2022
- 2022-12-05 CN CN202211547468.3A patent/CN115618749B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN115618749A (en) | 2023-01-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101246590B (en) | Star loaded camera spacing deformation image geometric correction method | |
CN106127697B (en) | EO-1 hyperion geometric correction method is imaged in unmanned aerial vehicle onboard | |
CN102741706B (en) | The geographical method with reference to image-region | |
CN105352509B (en) | Unmanned plane motion target tracking and localization method under geography information space-time restriction | |
CN106871927B (en) | Unmanned aerial vehicle photoelectric pod installation error calibration method | |
US7071970B2 (en) | Video augmented orientation sensor | |
CN101598556B (en) | Unmanned aerial vehicle vision/inertia integrated navigation method in unknown environment | |
CN103345737B (en) | A kind of UAV high resolution image geometric correction method based on error compensation | |
CN109708649B (en) | Attitude determination method and system for remote sensing satellite | |
CN106780321B (en) | CBERS-02 satellite HR sensor image overall tight orientation and correction splicing method | |
CN103822615A (en) | Unmanned aerial vehicle ground target real-time positioning method with automatic extraction and gathering of multiple control points | |
CN107040695B (en) | satellite-borne video image stabilization method and system based on RPC positioning model | |
CN115618749B (en) | Error compensation method for real-time positioning of large unmanned aerial vehicle | |
CN114858133B (en) | Attitude low-frequency error correction method under fixed star observation mode | |
CN114111776A (en) | Positioning method and related device | |
CN113415433A (en) | Pod attitude correction method and device based on three-dimensional scene model and unmanned aerial vehicle | |
CN113739767B (en) | Method for producing orthoscopic image aiming at image acquired by domestic area array swinging imaging system | |
CN110411449B (en) | Aviation reconnaissance load target positioning method and system and terminal equipment | |
CN107705272A (en) | A kind of high-precision geometric correction method of aerial image | |
CN113654528B (en) | Method and system for estimating target coordinates through unmanned aerial vehicle position and cradle head angle | |
Li et al. | Multi-sensor based high-precision direct georeferencing of medium-altitude unmanned aerial vehicle images | |
CN113049006B (en) | Starlight vector Mongolian difference correction method based on model parameter estimation | |
CN111044076B (en) | Geometric calibration method for high-resolution first-number B satellite based on reference base map | |
CN112257630A (en) | Unmanned aerial vehicle detection imaging method and device of power system | |
CN117990112B (en) | Unmanned aerial vehicle photoelectric platform target positioning method based on robust unscented Kalman filtering |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |